Large simulated datasets of lateral (key) pinch. The data provided represent shifts in thumb-tip force associated with scaling a musculoskeletal model of the wrist and thumb, as well as isolated adjustments to Hill-type parameters.
This work elucidates the biomechanics of the thumb during lateral (i.e. key) pinch (a clinical outcome measure and activity of daily living) under a variety of conditions. Using a previously published wrist-thumb model (Nichols et al., 2017), we forward dynamically simulated large datasets of lateral pinch. These datasets display the impact of anthropometrically scaling the wrist-thumb model, as well as isolated variations in Hill-type muscle parameters.
We have used the included datasets to determine the efficacy of using anthropometrically scaled generic models to capture age-dependent differences in lateral pinch force. The ages investigated include childhood, puberty, older adolescence, and adulthood. Simulated muscle activations (from computed muscle control) and lateral pinch forces (from forward dynamics) were compared against those from existing literature. While anthropometric scaling could capture variations in lateral pinch force, a generic muscle control strategy is not representative of all populations.
Motivated by the difficulties of accurate muscle-tendon parameter selection, we also tested the ability of artificial neural networks to classify a Hill-type muscle parameter from lateral pinch force alone. This work used large, dynamic datasets of lateral pinch force to elucidate the impact of altering the maximum isometric force of thumb muscle actuators. The size of the datasets ranged from 120 to 4096 simulations, corresponding to the adjustment of additional thumb muscles. This work demonstrated that artificial neural networks may be an inexpensive approach for approximating Hill-type muscle parameters. We also identified that including muscles with redundant function may decrease machine learning model accuracy.